Capital Intensity as the Primary Moat in AI Investing

Sameer Dholakia of Bessemer Venture Partners describes a systemic shift toward an intelligence supercycle, suggesting the current AI era represents a market larger than oil. His core thesis is that capital has become a primary moat in a compute-intensive landscape. This forces a professionalization of venture capital that favors either massive, global platforms or extreme specialists. For founders and investors, the consequence is that traditional Triple, Triple, Double, Double growth metrics no longer secure capital in a world where hyper-growth is the new baseline. Readers gain an advantage by understanding how Dholakia distinguishes between option bets and generational companies, providing a framework for navigating the high-stakes, capital-intensive reality of modern AI investing.

The Hidden Cost of Capital-Intensive Moats

In the current AI landscape, the competitive environment is defined by the capital intensity required to enter the game. Dholakia notes that while the market for AI applications is vast, the infrastructure layer, which he calls the heartbeat of intelligence, is massively compute-intensive.

> It turns out that in the market we were looking at here, he was going to be the same thing. It was massively compute intensive and therefore capital intensive. And so it is going to consolidate to a handful of players.
>
> -- Sameer Dholakia

This creates a systemic feedback loop: capital acts as a moat, which attracts top-tier talent, which in turn reinforces the company market position. For the investor, this shifts the focus from traditional valuation metrics to the ability to back up the truck once a winner emerges. The immediate discomfort of high entry prices is a necessary trade-off for capturing a generational asset.

Why the Obvious Fix Makes Things Worse

Conventional wisdom suggests that open-source models will commoditize the market and hurt the business models of frontier providers. Dholakia challenges this, calling it a tyranny of the or fallacy. He argues that the market for intelligence is so vast that the mix between closed and open tokens is secondary to the sheer scale of the total addressable market.

> I really do believe that there are all those different use cases from consumers up through enterprise and government, and they are all going to want intelligence of different types and different flavors.
>
> -- Sameer Dholakia

This perspective reveals a non-obvious dynamic: the system routes around the closed versus open debate. Enterprises will use both, creating room for multiple winners. The downstream effect is that investors who optimize for a specific mix of models miss the broader opportunity, while those who maintain exposure to both ends of the stack capture the systemic growth.

The 18-Month Payoff Nobody Wants to Wait For

When evaluating high-growth AI companies, Dholakia ignores the rear-view mirror of current ARR multiples, which he admits would make most traditional investors throw up. Instead, he anchors his analysis on the next 12 to 18 months of compounding growth. This requires a level of patience and conviction that most market participants lack.

The system rewards those who can look past the current valuation to the 5-year growth trajectory. As Dholakia observes, businesses like Legora or Anthropic were considered insane at their entry valuations, yet they demonstrated that in a supercycle, the multiple paid today is often dwarfed by the revenue scale achieved within two years. The competitive advantage here is purely temporal: most investors cannot justify the discomfort of the entry price, leaving the field open to those who can model the 18-month inflection point.

Key Action Items

• Audit your Option Bets (Immediate): Distinguish between companies that are truly generational and those that are merely option bets. If you are an entrepreneur, accept that you must earn the right to the next round of capital; if you are an investor, ensure your reserve strategy allows you to double down on the winners.
• Evaluate TAM through a Global Lens (Next Quarter): Stop limiting your market analysis to domestic borders. AI and intelligence-driven businesses, like those in legal or auditing tech, have global ambitions. If your TAM calculation does not account for international service dollars, you are likely underestimating the ceiling.
• Prioritize Teammates over Colleagues (12-18 Months): When building your firm or portfolio companies, move beyond colleagues. Create systemic opportunities for non-work interaction, such as kayaking or mountain biking, to build the social capital required to survive when things go wrong.
• Shift to Next 12 Months (NTM) Valuation Models (Immediate): Stop anchoring on current ARR multiples. In hyper-growth AI, NTM multiples of 10-15x are reasonable if the company is blowing through its growth targets.
• Prepare for the IPO Unlock (12-24 Months): Anticipate a reopening of the IPO window by 2028. Use this time to build IPO-worthy operational discipline, as public markets will demand pure-play AI exposure and high-rigor reporting.